Prognostic factors in patients receiving third-line targeted therapy for metastatic renal cell carcinoma.

Transcription

1 Prognostic factors in patients receiving third-line targeted therapy for metastatic renal cell carcinoma. Iacovelli R 1, Farcomeni A 2, Sternberg CN 3, Carteni G 4, Milella M 5, Santoni M 6, Cerbone L 3, Di Lorenzo G 7, Verzoni E 1, Ortega C 8, Sabbatini R 9, Ricotta R 10, Messina C 11, Lorusso V 12, Atzori F 13, De Vincenzo F 14, Sacco C 15, Boccardo F 16, Valduga F 17, Massari F 18, Baldazzi V 19, Cinieri S 20, Mosca A 21, Ruggeri EM 22, Berruti A 23, Procopio G 1 1. Medical Oncology Unit 1, Fondazione IRCCS Istituto Nazionale dei Tumori, Milan, Italy. 2. Department of Public Health and Infectious Diseases, Sapienza-University of Rome, Rome, Italy. 3. Department of Medical Oncology, San Camillo Forlanini Hospital, Rome, Italy. 4. Oncology Unit, A. Cardarelli Hospital, Naples, Italy. 5. Medical Oncology A, Regina Elena National Cancer Institute, Rome, Italy. 6. Department of Medical Oncology, Polytechnic University of the Marche Region, Ancona, Italy. 7. Medical Oncology, Genitourinary Cancer Section, University Federico II, Naples, Italy. 8. Fondazione del Piemonte per l'oncologia IRCC, Candiolo, Italy; 9. Oncology Division, Department of Oncology and Hematology, University of Modena e Reggio Emilia, Modena, Italy. 10. The Falck Division of Oncology, Ospedale Niguarda Ca Granda, Milan, Italy. 11. Oncology Division, Ospedali Riuniti, Bergamo, Italy; 12. National Cancer Research Center, Istituto Tumori Giovanni Paolo II Bari, Italy; 13. Medical Oncology Unit, Azienda Ospedaliero Universitaria of Cagliari, Italy; 14. Oncology and Hematology Unit, Humanitas Cancer Center, Istituto Clinico Humanitas, Rozzano, Italy. 15. Oncology Unit, St Maria della Misericordia Hospital, Udine, Italy; 16. University and IRCCS AOU-San Martino-IST, National Cancer Research Institute, Genoa, Italy; 17. Medical Oncology, St. Chiara Hospital, Trento, Italy. 18. Medical Oncology, G.B. Rossi Academic Hospital, University of Verona, Verona, Italy; 19. Department of Medical Oncology; Santa Maria Annunziata Hospital, Florence, Italy; 20. Medical Oncology & Breast Unit Department, Sen A. Perrino Hospital, Brindisi, Italy; 21. Medical Oncology, Maggiore della Carità University Hospital, Novara, Italy; 22. Oncology Unit, Belcolle Hospital, Viterbo, Italy;

2 23. Medical Oncology, University of Brescia, Brescia, Italy; Corresponding Author: Dr. Roberto Iacovelli Istituto Nazionale dei Tumori di Milano Via G Venezian Milan; Italy Tel: roberto.iacovelli@istitutotumori.mi.it

3 ABSTRACT Purpose: Several prognostic models have been proposed for metastatic renal cell carcinoma (mrcc) but none has been validated in patients receiving third-line targeted agents. We aim to evaluate the prognostic factors in patients affected by mrcc who received a third-line of targeted agent. Materials and Methods: Data about 2,065 patients affected by clear cell mrcc and treated with targeted therapies in 23 centres in Italy were retrospectively reviewed. A total of 281 patients treated with three targeted agents were included in the final analysis. Overall survival (OS) was the main outcome. Cox proportional hazards regression, followed by bootstrap validation, was used to identify independent prognostic factors. Results: Three clinical characteristics (ECOG-PS >1, presence of metastases at diagnosis, and presence of liver metastases) and two biochemical factors (hemoglobin < LLN and neutrophils > ULN) were found to be prognostic. Patients were classified in three categories of risk based on the presence of zero or one risk factors (low-risk), two risk factors (intermediate-risk) and >2 risk factors (high-risk). The median OS was 19.7, 10.1, and 5.5 months, while the 1-year OS was 71%; 43%, and 15%, respectively. Major limit is the retrospective nature of this study and the absence of external validation. Conclusions: This nomogram included both clinical and biochemical prognostic factors and it may be useful for selection of patients in clinical trials and for the defining prognosis in clinical practice. Key Words: mrcc; prognosis; third-line; targeted therapy; overall survival. Running Head: 3 rd line & prognosis in mrcc patients

4 Introduction Treatment of metastatic renal cell carcinoma (mrcc) has dramatically changed since targeted agents have been introduced in the clinical armamentarium. Despite the low curative rate, the median overall survival has increased over the years: from 10 months in the era of cytokines to about 30 months in the era of targeted therapies 1-3. The increased survival and the growing number of available targeted agents have resulted in a greater number of patients treated with two or more lines of therapy. Considering this wealth of new agents, identification of prognostic factors remains a cornerstone for clinical management of advanced disease. Prognostic factors allow stratification of patients based on cancer-related risk of death, and give important information about disease evolution. Moreover, this allow homogeneous stratification of patients for clinical trials in the attempt to avoid bias related to selection and consequently, identify the group in which a target agent has the greatest activity. Until now, the most-frequently used prognostic criteria were those elaborated by Motzer in patients treated with interferon immunotherapy and chemotherapy at Memorial Sloan-Kettering Cancer Centre (MSKCC) 4, and subsequently validated in a retrospective analysis on 353 patients treated at the Cleveland Clinic 5. Subsequently, Heng et al. reported a large analysis performed on patients treated with tyrosine kinase inhibitors (TKIs) as up-front or second-line therapy and included in the International mrcc Database Consortium (IMDC) 6. The role of low hemoglobin and high serum corrected calcium values, such as Karnofsky PS and the time from diagnosis to initiation of therapy were confirmed as independent predictors of short survival. Furthermore, the value of neutrophils and platelet count greater than the upper limit of normal (ULN) were also found to be prognostic 6. Similar to previous analysis, there were no prognostic differences if patients received targeted agents as first-line or after cytokines; moreover the role of histology was not investigated 7.

5 Despite their validation in therapy naïve patients, the MSKCC model has also been used in patients enrolled in second-line trials and, more recently, the IMDC model has confirmed its discriminatory capacity also in this setting 8. This study sought to evaluate prognostic factors in patients with mrcc who receive a third-line of targeted agent and to compare these factors with current nomograms. Patients and Methods Patients Data of 2,065 patients with mrcc and treated with targeted therapies in 23 centres in Italy were retrospectively reviewed. Only patients receiving three lines of targeted agents were included in the final analysis. Patient inclusion criteria were: diagnosis of metastatic clear cell RCC and treatment with three targeted therapies, while patients treated with a combination of therapies or previous cytokines were excluded. Baseline demographic, clinical and laboratory data and characteristics previously found to have prognostic value were collected retrospectively by using uniform database templates to ensure consistent data collection. Outcome data on overall survival (OS) were collected from patient files and by telephone contact. The study received Internal review board approval. Statistical Analyses The primary outcome was OS, defined as the time from initiation of third line therapy to death as a result of any cause or was censored at the date of last follow-up. The median OS with 95% CIs was estimated by using the Kaplan-Meier method. Associations between OS and potential prognostic factors were assessed by using the log-rank test in univariable analysis. P-values at univariable analysis were adjusted for multiplicity using the Bonferroni correction. The Cox proportional hazards model was then fitted in multivariate analyses. Model selection was performed using a

6 forward step-wise procedure. The proportionality of hazards assumption was assessed graphically by using plots of log (log[survival]) versus log of survival time. Once the prognostic factors were identified and the final model was fit, a risk-group variable was created by counting the number of unfavourable features for each patient. The predictive performance of the newly constructed score(s) was assessed by means of the C- index, which corresponds to the area under the ROC curve and represents the ability of a score to correctly predict events. A concordance index of 1 represents perfect ability to distinguish patients; an index > 0.5 implies good prediction ability, while index of 0.5 implies no predictive ability. We also assessed the predictive performance of the final model by internal validation by using twostep bootstrap resampling procedures. In the first step, 1000 bootstrap samples were generated randomly with replacement from the original study population. The stepwise Cox regression procedure was employed to each sample with the same selection criteria as the original modeling, described earlier in this methods section. We then calculated the frequency of inclusion of each variable in the resulting models within the 1000 bootstrap samples. Risk factors that were present in more than 50% of the models were considered significant. In the second validation step, we validated parameter estimates of the final model. One thousand bootstrap samples were generated randomly from the original study population for the final model. For each of the samples, we refit the Cox regression model by using the variables selected in the final model, and we calculated the regression parameters and hazard ratios. The means, standard deviations, and CIs were computed from the 1000 samples and were compared to the model by using the original study population. Improvement in comparison to other scores was assessed by means of continuous Net Reclassification Index (NRI) and median improvement in risk scores (MIRS), which were computed as described by Pencina et al. 9 The NRI gives roughly the proportion of misclassified cases that are classified correctly with the new information, a NRI of 1 indicates perfect ability to correctly reclassify patients, and a negative NRI indicates that the score is worse than its competitor.

7 The NRI was evaluated at 1 year, and a sensitivity analysis comparing with 6 and 18 months revealed no dependence on the cut-off chosen. The MIRS gives the median increase in score for the cases, a large positive MIRS indicates that the predicted risk for patients with events is increased on average, while a large negative MIRS indicates that the predicted risk is decreased. All statistical analyses were undertaken by using R, and P <.05 (two sided) was considered statistically significant. Results Patients After the screening, 281 patients treated from August 2006 to June 2011were included in the final analysis. The baseline patient characteristics are reported in Table 1. At the cut-off time, 73.7% of patients had progression on targeted therapy and 26.3% had ongoing third line treatment. A total of 46.6% of patients are dead at the time of the analysis, the median follow-up is 8 months and the maximum 42 months. The median overall survival for the entire cohort was 13.8 months (95%CI, ) and the 1-year overall survival was 54% (95%CI, 47.6% 61.6%) (Figure 1). Univariable Analysis All baseline characteristics were tested for predictive value at univariable analysis as reported in Table 2. The predictive variables associated with poor overall survival were ECOG PS, metastatic disease at diagnosis, hemoglobin lower than the LLN, value of ULN of neutrophils, LDH greater than the ULN, two or more sites of metastases and the presence of hepatic or bone metastases. Multivariable Analysis The predictive variables associated with poor overall survival at multivariable analysis were ECOG PS, the presence of metastases at diagnosis, age greater than 60 years, hemoglobin lower the LLN,

8 ULN of neutrophils and the presence of liver metastases. Some of these factors have been validated in other classifications such as the MSKCC, the IMDC and the French models. We employed five of these prognostic factors to classify patients into three categories and in the Italian RCC Third-Line (IRTL) prognostic score, while age greater than 60 years was excluded due to the mild effect on prognosis (e.g. reduction of 2% of the risk of death) (Table 3). Patients with zero or one risk factor were classified as low-risk; patients with two risk factors were classified as intermediate-risk and patients with more than two risk factors were classified as high-risk; sixteen patients were not classified. The 153 patients with low-risk had a median OS of 19.7 months (95%CI, ) and a 1-year OS of 71%; 67 patients with intermediate-risk had a median OS of 10.1 months (95%CI, ) and a 1-year OS of 43%, and the 45 patients with high-risk had a median OS of 5.5 months (95%CI, ), and a 1-year OS of 15% (Figure 2). The overall HR comparing high vs. intermediate risk was 2.20 (95%CI, ; p=0.0011); while intermediate vs. low risk had an HR of 2.47 (95% CI, , p<0.0001). Bootstrap Validation The stepwise Cox regression procedure was employed with each of the 1,000 random bootstrap samples with the same selection criteria as the original modeling. The regression parameters and hazard ratios produced from the 1,000 bootstrap samples were similar to the original model, which suggests excellent internal validation (Supplementary Table 1). The bootstrap biased-corrected C-index of this model was (95% CI, ). Validation of previous prognostic nomograms We performed a sensitivity analysis to identify the prognostic role of nomograms validated in patients with mrcc treated with first-line therapy. The distribution of patients and the median overall survivals based on prognostic group for each nomogram was reported in Table 4.

9 In the MSKCC model the HR between the good and intermediate group was 2.41 (p<0.001) and between the intermediate and poor group was 2.66 (p<0.001). In the CCF model the HR between the good and intermediate group was 3.21 (p<0.001) and between the intermediate and poor group was 1.51 (p<0.001). In the French model the HR between the good and intermediate group was 2.44 (p=0.080) and between the intermediate and poor group was 2.25 (p=0.002). In the IMDC model the HR between the good and intermediate group was 1.54 (p=0.004) and between the intermediate and poor group was 3.19 (p<0.001). C-indexes between IRTL prognostic score and other prognostic classifications are reported in Supplementary Table 2. The data reported in previous models may be equally effective to predict prognosis for third line therapy. Therefore, we decided to compare our model with the previous ones both comparing the prognostic factors and the prognostic categories using the net-reclassification-index (NRI) and the Median Improvement in Risk Score (MIRS). Both analyses suggest that the IRTL prognostic score may be useful compared to MSKCC, CCF, French and IMDC prognostic models in defining the prognosis of these patients both in terms of prognostic factors and prognostic categories (Supplementary Table 3). Discussion In this article, we assess clinical factors associated with prognosis in mrcc treated with two previous lines of targeted agents and eligible for a third one. Prognostic factors have been widely investigated in treatment naïve mrcc patients and used to select patients for clinical trials. The first model analyzing 670 patients enrolled in clinical trials with immunotherapy and chemotherapy at MSKCC, found that hemoglobin, lactate dehydrogenase, corrected calcium, nephrectomy, and Karnofsky PS were independent risk factors for survival 4. Ten years later, Heng et al. confirmed the prognostic role of low hemoglobin, high serum corrected calcium, low Karnofsky PS and time from diagnosis to therapy less than 1 year and added the absolute value of neutrophils and platelets counts greater than ULN 6. The study reported a median OS of 43.2 months

10 in the favorable risk group, 22.5 months in the intermediate risk, and 7.8 months in the poor risk group. This model was also externally validated on more than 840 patients; it was found as IMDC has a good concordance index (c-index: 0.66) and is able in reclassifying patients more correctly than the majority of other models 10. Our model, similarly to others, was able to confirm the prognostic role for some clinical and biochemical factors previously identified by Motzer and Heng also in the third-line setting and added the presence of liver metastases. Our analysis confirms that different prognostic groups of patients may be found before first-line as well as before third-line where the median life-expectancy is generally shorter. About further lines, the literature reports that 50% of patients receive two lines of therapy and about 15% will receive three lines 11, 12. It has been recently shown that patients who are able to receive several lines of therapy survive longer. Yet, there are no predictive tools to select who will be eligible for further lines 13. The final OS reported by the population included in this analysis (i.e. 13 months), is comparable with results of a prospective phase III trial comparing dovitinib to sorafenib in the third-line setting 14. This long survival may suggest that prediction of prognosis still remains an important factor because low-risk patients have a median OS increased by three times as compared to high-risk patients. Our analysis also reports the clinical characteristics of patients who received third-line therapy: as for naïve patients, major site of metastases were lung, followed by lymph-nodes, bones and liver 15. Despite the high number of baseline clinical characteristics investigated, the most important prognostic factors were comparable to previous classifications such as PS, platelet count and neutrophils count. Differently from the MSKCC and IMDC models, the present analysis reports the independent prognostic role of liver metastases and includes this factor in the final model, similar to the French model 16. This factor was also investigated in patients included in the IMDC model revealing that 19% of patients had liver metastases at the beginning of first-line therapy, confirming its negative

11 prognostic role also when compared to other prognostic factors 17. This data suggest that the presence of liver metastases is an independent prognostic factor in patients treated with targeted agents over different lines of therapy and probably, it worsened prognosis in subsequent lines. When the IRTL model was compared to other nomograms using the NRI and MIRS tests, these suggest a better performance, especially when the comparison was based on prognostic factors more than when based on prognostic categories. Our study has some limitations: first and foremost is its retrospective nature that did not ensure a complete data collection for all patients. Moreover, we included only patients with clear cell histology and no data are available on papillary or chromophobe renal tumors. The use of other prognostic score not designed and not yet validated in patients treated with two previous lines of therapy may affect the comparison with a score specifically designed for this population. On the other hand, previous prognostic model have been evaluated and validated on larger patient populations compared to those reported in this study. Nevertheless, considering the low number of patients receiving three lines of therapy this nomogram may be considered a useful tool to study this population. Conclusions The IRTL prognostic score included three clinical prognostic factors such as the ECOG PS, the presence of metastases at diagnosis, the presence of liver metastases and two biochemical ones; such as hemoglobin below the LLN and the neutrophils count above the ULN. The discriminatory power among prognostic groups indicates that this model may be useful for selection of patients for clinical trials and for the definition of prognosis in clinical practice even if an external validation is recommended. Author Disclosures Authors have not conflict of interest related to this work

12 References: 1. Pa SK, Figlin RA. Targeted therapies for renal cell carcinoma: understanding their impact on survival. Target Oncol Jun;5(2): Motzer RJ, Hutson TE, McCann L, et al. Overall survival in renal-cell carcinoma with pazopanib versus sunitinib. N Engl J Med. 2014; 370: Iacovelli R, Alesini D, Palazzo A, et al. Targeted therapies and complete responses in first line treatment of metastatic renal cell carcinoma. A meta-analysis of published trials. Cancer Treat Rev. 2014; 40: Motzer RJ, Mazumdar M, Bacik J, et al. Survival and prognostic stratification of 670 patients with advanced renal cell carcinoma. J Clin Oncol. 1999; 17: Mekhail TM, Abou-Jawde RM, Boumerhi G, et al. Validation and extension of the Memorial Sloan-Kettering prognostic factors model for survival in patients with previously untreated metastatic renal cell carcinoma. J Clin Oncol. 2005; 23: Heng DY, Xie W, Regan MM, et al. Prognostic factors for overall survival in patients with metastatic renal cell carcinoma treated with vascular endothelial growth factor-targeted agents: results from a large, multicenter study. J Clin Oncol. 2009; 27: Choueiri TK, Lim ZD, Hirsch MS, et al. Vascular endothelial growth factor-targeted therapy for the treatment of adult metastatic Xp11.2 translocation renal cell carcinoma. Cancer. 2010; 116: Ko JJ, Xie W, Heng DY, et al. The International Metastatic Renal Cell Carcinoma Database Consortium (IMDC) model as a prognostic tool in metastatic renal cell carcinoma (mrcc) patients previously treated with first-line targeted therapy (TT). J Clin Oncol 32, 2014 (suppl 4; abstr 398). 9. Pencina MJ, D'Agostino RB, Steyerberg EW. Extensions of net reclassification improvement calculations to measure usefulness of new biomarkers. Stat Med. 2011; 30: Heng DY, Xie W, Regan MM, et al. External validation and comparison with other models of the International Metastatic Renal-Cell Carcinoma Database Consortium prognostic model: a population-based study. Lancet Oncol. 2013; 14: Iacovelli R, Cartenì G, Sternberg CN, et al. Clinical outcomes in patients receiving three lines of targeted therapy for metastatic renal cell carcinoma: results from a large patient cohort. Eur J Cancer. 2013; 49: Levy A, Menard J, Albiges L, et al. Second line treatment of metastatic renal cell carcinoma: The Institut Gustave Roussy experience with targeted therapies in 251 consecutive patients. Eur J Cancer. 2013; 49: Ko JJ, Choueiri TK, Rini BI, et al. First-, second-, third-line therapy for mrcc: benchmarks for trial design from the IMDC. Br J Cancer. 2014; 110:

13 14. Motzer RJ, Porta C, Vogelzang NJ, et al. Dovitinib versus sorafenib for third-line targeted treatment of patients with metastatic renal cell carcinoma: an open-label, randomised phase 3 trial. Lancet Oncol. 2014; 15: Bianchi M, Sun M, Jeldres C, et al. Distribution of metastatic sites in renal cell carcinoma: a population-based analysis. Ann Oncol. 2012; 23: Négrier S, Escudier B, Gomez F, et al. Prognostic factors of survival and rapid progression in 782 patients with metastatic renal carcinomas treated by cytokines: a report from the Groupe Français d'immunothérapie. Ann Oncol. 2002; 13: McKay RR, Kroeger N, Xie W, et al. Impact of bone and liver metastases on patients with renal cell carcinoma treated with targeted therapy. Eur Urol. 2014; 65:

14 Legend of the Figures: Figure 1: Final overall survival. Figure 2: Overall survival by IRTL prognostic group.

15 Table 1: Baseline characteristics of the patients. Variable Patients (%) Median age, years (IQR) 62.6 (15.1) Male sex 74.7 Nephrectomy 77.2 Metastatic at diagnosis 37.7 Interval < 1 year 50.2 First line targeted agents Sunitinib 62.5 Sorafenib 22.9 Bevacizumab+ IFN 9.6 Temsirolimus 2.5 Other 2.5 Second line targeted agents Sunitinib 28.1 Sorafenib 33.1 Everolimus 27.8 Temsirolimus 10.0 Other 1.0 Third line targeted agents Sunitinib 7.5 Sorafenib 32.5 Everolimus 51.1 Temsirolimus 5.0 Other 3.5 ECOG Performance Status Previous RT 34.4 Sites of metastases Sites of metastases Lung 77.3 Bone 32.1 Liver 28.6 Adrenal 21.7 Brain 12.3 Pancreatic 9.4 Lower lymphnode 24.3 Upper lymphnode 34.8 Soft tissue 14.9 Hb < LLN 46.3 LDH > 1.5 ULN 20.1 Neu > ULN 10.8 PLT > ULN 7.5 Corrected Ca 2+ > ULN 3.6

16 Table 2: Univariate analysis based on patient characteristics before the third-line of therapy. The p- values at univariablee analysis are Bonferroni adjusted for multiplicity. Variable Univariate Analysis Bootstrap validation HR 95% CIs p-value % times selected Nephrectomy (Y/N) Metastatic at diagnosis (Y/N) Sex (M/F) Age > 60 years ECOG Performance Status (0/1 vs. 2) < Interval < 1 year (Y/N) Previous RT (Y/N) Hb < LLN (Y/N) < LHD > 1.5 ULN (Y/N) Neu > ULN (Y/N) < PLT > ULN (Y/N) Corrected Ca 2+ > ULN (Y/N) Sites of metastases 2 (Y/N) Lung metastases (Y/N) Bone metastases (Y/N) < Liver metastases (Y/N) Adrenal metastases (Y/N) Brain metastases (Y/N) Pancreatic metastases (Y/N) Lower lymphnode metastases (Y/N) Upper lymphnode metastases (Y/N) Soft tissues metastases (Y/N) HR= hazard ratio; CIs= confidence intervals; Y= tes; N= non; M= male ; F= female ; Hb= hemoglobin ; Neu= neutrophils ; LLN= low limit of normal ; ULN= upper limit of normal; PLT= platelets; LDH= lactate dehydrogenase.

17 Table 3: Results of Multivariate Analysis. Variable Multivariate analysis HR 95% CIs p-value Metastatic at diagnosis (Y/N) < ECOG Performance Status 0/1 vs <0.001 Hb < LLN (Y/N) Neu > ULN (Y/N) Liver metastases (Y/N) HR= hazard ratio; CIs= confidence intervals; SD= standard deviation; Y= yes; N= non; Hb= hemoglobin; Neu=neutrophils; ECOG= Eastern Cooperative Oncology Group.

18 Table 4: Distribution of patients based on prognostic nomogram. Prognostic Nomograms: Prognostic Group MSKCC French CCF IMDC (%) Median OS (95%CIs) (%) Median OS (95%CIs) (%) Median OS (95%CIs) (%) Median OS (95%CIs) Good 20.6 NR (20.3 NA) 5.3 NR (11.1 NA) 20.3 NR (NR NR) (13.2 NR) Intermediate ( ) ( ) ( ) ( ) Poor ( ) (4.7 NR) ( ) (3.0 NR) MSKCC= Memorial Sloan Kettering Cancer centre; CCF= Cleveland Clinic Foundation; IMDC= International mrcc Database Consortium; OS= overall survival; CIs= Confidence Intervals; NR= not reached.

19 Supplementary table 1: Results of bootstrap validation of the multivariate model. Variable HR Bootstrap means 95% CIs SD of HR Metastatic at diagnosis (Y/N) ECOG Performance Status 0/1 vs Hb < LLN (Y/N) Neu > ULN (Y/N) Liver metastases (Y/N) HR= hazard ratio; CIs= confidence intervals; SD= standard deviation; Y= yes; N= non; Hb= hemoglobin;neu=neutrophils; ECOG= Eastern Cooperative Oncology Group.

20 Supplementary table 2: C-indexes based on comparison of the Italian RCC Third Line (IRTL) prognostic model and other prognostic nomograms. Prognostic Nomogram C-index 95% CIs p-value MSKCC CCF French <0.001 IMDC <0.001 MSKCC=memorial Sloan Kettering cancer centre; CCF=Cleveland clinic Foundation; IMDC=International mrcc Database Consortium.

21 Supplementary Table 3: Comparison between Italian RCC Third Line (IRTL) prognostic model and other prognostic models in mrcc. Comparison was based on numbers of prognostic factors (prognostic factors) or based on categories of risk found in each model (prognostic score). Prognostic models IRTL prognostic model Prognostic factors Prognostic categories Median (95%CIs) p-value Median (95%CIs) p-value MSKCC NRI 0.29 ( ) ( ) MIRS 0.05 ( ) ( ) CCF NRI 0.25 ( ) ( ) MIRS 0.10 ( ) ( ) French NRI 0.27 ( ) ( ) MIRS 0.16 ( ) < ( ) <0.001 IMDC NRI 0.25 ( ) ( ) MIRS 0.02 ( ) ( ) MSKCC= Memorial Sloan Kettering Cancer centre; CCF= Cleveland Clinic Foundation; IMDC= International mrcc Database Consortium; NRI= net reclassification index; MIRS= Median Improvement in Risk Score; CIs= Confidence Intervals.